PSFC/JA-10-42 A unified treatment of kinetic effects in a tokamak pedestal Peter J. Catto, Gregory Kagan1, Matt Landreman, and Istvan Pusztai2 1 Theoretical Division, Los Alamos National Laboratory, Los Alamos, NM 87545, USA Nuclear Engineering, Applied Physics, Chalmers University of Technology and Euratom-VR Association, SE-41296 Goteborg, Sweden 2 November 2010 Plasma Science and Fusion Center Massachusetts Institute of Technology Cambridge MA 02139 USA This work was supported by the U.S. Department of Energy, Grant No. DE-FG02-91ER54109. Reproduction, translation, publication, use and disposal, in whole or in part, by or for the United States government is permitted. Submitted to Plasma Phys. Control. Fusion (August 2010) A unified treatment of kinetic effects in a tokamak pedestal Peter J. Catto1, Grigory Kagan2, Matt Landreman1, Istvan Pusztai3 1) Plasma Science and Fusion Center, Massachusetts Institute of Technology, Cambridge, MA 02139, USA 2) Theoretical Division, Los Alamos National Laboratory, Los Alamos, NM 87545, USA 3) Nuclear Engineering, Applied Physics, Chalmers University of Technology and EuratomVR Association, SE-41296 Goteborg, Sweden e-mail contact: catto@psfc.mit.edu Abstract We consider the effects of a finite pedestal radial electric field on ion orbits using a unified approach. We then employ these modified orbit results to retain finite E×B drift departures from flux surfaces in an improved drift kinetic equation. The procedure allows us to make clear distinction between transit averages and flux surface averages when solving this kinetic equation. The technique outlined here is intended to clarify and unify recent evaluations of the banana regime decrease and plateau regime alterations in the ion heat diffusivity, the reduction and possible reversal of the poloidal flow in the banana regime and its augmentation in the plateau regime, the increase in the bootstrap current, and the enhancement of the residual zonal flow regulation of turbulence. 1. Introduction It is standard in tokamak kinetic theory to assume the poloidal ion gyroradius is small compared to all macroscopic scale-lengths, but in the pedestal this approximation can break down. This challenge is addressed by our recently developed techniques that account for the presence of short radial scale lengths in a subsonic pedestal. Using these techniques, we have calculated both the banana [1,2] and plateau [3] regime modifications to the neoclassical ion heat flux, ion and impurity flows, the bootstrap current; and extended the residual zonal flow calculation to the pedestal case [4,5]. Here, we present a general formalism providing deeper insight into these detailed calculations by considering the effects of the finite pedestal radial electric field E on the ion orbits by an improved procedure. This unified approach highlights intricacies of the previous investigations, thereby forming a firm basis for how these kinetic calculations are best performed. In particular, we present a general method of treating finite electric field effects on ion orbits that allows us to solve a reduced kinetic equation for the pedestal modifications to the results of the usual evaluations of neoclassical transport in the banana and plateau regimes [6-8], and the residual zonal flow calculation [9,10]. The emphasis herein is on the steps that differ from the usual neoclassical and residual zonal flow treatments. The basis of the technique is the convenience of employing the canonical angular momentum as the radial gyrokinetic variable [11]. This choice allows us to perform calculations in the pedestal by accounting for the presence of strong radial density (and electron temperature) gradient scale length on the order of a poloidal ion gyroradius ρpi = ρiB/Bp, where B/Bp >> 1 and ρ i = vi/Ωi is the ion gyroradius with vi = (2Ti/M)1/2 the ion thermal speed and Ωi = ZeB/Mc the ion gyrofrequency for ions of mass M, charge number Z, and temperature Ti. In a subsonic pedestal the E×B drift and the ion diamagnetic drift must cancel to lowest order [11-13]. In the helium discharges on DIII-D this behavior is verified since the background ion temperature can be measured directly [14]. In this situation the ions are electrostatically confined to lowest order and the associated radial electric field is so large that the E×B drift velocity can compete with the poloidal component of the parallel streaming to modify trajectories. These modifications introduce electric field dependence into the neoclassical ion and impurity flows via the usual ion temperature gradient term [1,2]. The ion temperature pedestal is always at least B/Bp wider than ρpi because of the constraint that the entropy production must vanish in a banana or plateau regime pedestal [11]. In addition, the ion heat flux [1,3] and the bootstrap current [2,3] are altered, along with the residual zonal flow [4,5] response of Rosenbluth and Hinton [9]. In spite of these changes the plasma remains intrinsically ambipolar [15,16] in the more general sense that E (or more precisely, the difference between the E×B and ion diamagnetic drifts) remains undetermined until conservation of toroidal angular momentum is solved, but unlike the core, the heat flux does depend on the radial electric field as does the ambipolar particle flux. In section 2 we present a unified approach to retaining the effect of the radial electric field on the ion orbits to be used for both the banana and plateau regime cases. A drift kinetic equation valid for treating neoclassical transport in both regimes of collisionality that also retains the zonal flow residual is efficiently derived in section 3. The generality of the results in these two sections then allows us to conveniently derive all earlier neoclassical banana [1,2] and plateau [3] results in section 4, as well as all previous results for the radial electric field modified zonal flow residual [4] and its orbit squeezing generalization [5] in section 5. Moreover, the treatment of collisions is streamlined so that the transit averaged collisional constraints are performed retaining finite drift orbit effects. € 2. Radial electric field effects on the treatment of ion orbits When the radial density scale length of the pedestal becomes comparable to the poloidal ion gyroradius, ρpi, the ion flow speed can only be subsonic if the E × B and ion diamagnetic flows cancel to lowest order. The lowest order cancellation means that the ions are electrostatically confined with the radial electric field satisfying (Ze/Ti)dΦ/dr ≈ -dlnpi/dr € field results in an ~ 1/ ρ pi >> 1/a, with a the minor radius [11]. Such a strong radial electric E × B drift that competes with the poloidal component of parallel streaming while staying well below vi, since cE × B/B2 ~ (Bp/B)vi, where B = I∇ζ + ∇ζ × ∇ψ with ζ , ϑ and ψ the toroidal angle, poloidal angle and poloidal flux variables, respectively, B = B , and ∇ψ = € € € € € € € € RBp. For a dΦ/dr this large, the variation in potential energy over an orbit width is sufficient to make electrostatic trapping important, even when the potential is a flux function [1-5,11]. In addition to introducing this finite orbit effect, orbit squeezing can enter [17]. T h e competition between the E × B drift and the poloidal projection of parallel streaming makes it necessary to retain the distinction between the surfaces of constant magnetic flux and the drift surfaces on which the canonical angular momentum remains constant. However, to obtain analytical results,€we must assume the inverse aspect ratio ε is small compared to unity ( ε << 1). With this assumption the transport remains local because the trapped and barely passing ion orbit widths are of order ε1/ 2ρ pi , and so less than the equilibrium pedestal scale length that € € is allowed to be as small as ρ pi . 2 /2B, To see how the orbits are modified we employ the magnetic moment µ = v⊥ along with the drift€approximation to the canonical angular momentum (1) ψ∗ ≡ψ− (Mc/Ze)R 2 v ⋅∇ζ ≈ψ − (Iv|| /Ωi ) , € € energy variable since we consider Bp/B << 1, and the notationally convenient pseudo kinetic and constant of the motion €E ∗ ≡E − (Ze /M)Φ(ψ∗ ) = v 2 /2 + (Ze /M)[Φ(ψ) − Φ(ψ∗ )] , (2) 2 where E = v /2 + (Ze /M)Φ(ψ) is the total energy that is a constant of the motion. We also assume the potential only depends on ψ with a quadratic dependence so that using (1) gives € (3) Φ(ψ) = Φ(ψ∗ ) + (Iv|| / Ωi )Φʹ′(ψ∗ ) + (1 / 2)(Iv|| / Ωi )2 Φʹ′ʹ′ , € with Φʹ′ʹ′ a constant. The utility of these variables will become clearer in the next section € the drift kinetic equation. In this section we obtain the relations when they are used to rewrite needed in subsequent sections. We define the effective E × B velocities in the poloidal magnetic field and the orbit squeezing factor S as (4) u = cIΦʹ′/B , u∗ = cIΦ∗ʹ′/B , and S = 1+ (cI2Φʹ′ʹ′/Ωi B) , € with Φʹ′(ψ∗ ) = Φ∗ʹ′ and Φʹ′(ψ) = Φʹ′ . We then use Φʹ′(ψ) = Φʹ′(ψ∗ ) + (ψ − ψ∗ )Φʹ′ʹ′ to obtain the useful and important relation € € (5) v|| + u = Sv|| + u∗ . € In addition, we€can use the definitions (4) and the preceding result to rewrite E ∗ as E ∗ ≡Sv||2 / 2 + µB + u∗v|| = [(Sv|| + u∗ )2 / 2S] + µB − u∗2 /2S = [(v|| + u)2 / 2S] + µB − u∗2 /2S . (6) 2 1/2 Solving (6) for v|| gives v€ so sgn(v||+u) = sgn(Sv||+u*) || + u = Sv|| + u ∗ = ±(2SE ∗ + u ∗ − 2SµB) must also be specified when ψ*, ϑ, E *, µ are used as the variables.€Unlike the more familiar situation with ψ, ϑ, E, µ as the variables, sgn(v||) is no longer a useful coordinate since two phase space locations in ψ*, ϑ, E*, µ can have the same sign. Phase space remains split into trapped and passing regions defined by the preceding randicand at ϑ = π. The trapped distribution function must be independent of sgn(Sv||+u*) at the bounce points where it vanishes. This property is used in section 4 to define the new transit average annihilation operation holding fixed ψ*. The preceding results are valid for arbitrary aspect ratio. To make further progress it is necessary to introduce the inverse aspect ratio expansion of the magnetic field by letting B0/B = 1 - 2εsin2(ϑ/2) = 1 - ε + εcosϑ , (7) where B0 is the magnetic field at ϑ = 0. Retaining inverse aspect ratio corrections through order ε we may then write (6) as (8) (v|| + u) 2 /2 = (Sv|| + u∗ ) 2 /2 = W(1− ΛB/B0 ) = (1− Λ)W[1− κ 2 sin 2 (ϑ /2)] , 2 where W, Λ and κ are adiabatic invariants to order ε defined by 2 (9) W = S0E ∗ + 2(S0 −1)(E ∗ − µB0 ) + (3/2)u∗0 2 € S0µB0 + 2(S0 −1)(E ∗ − µB0 ) + u∗0 κ2 (10) Λ= 2 = 2 κ + 2ε S0E ∗ + 2(S0 −1)(E ∗ − µB0 ) + (3/2)u∗0 2 ] € 2εΛ 2ε[S0µB0 + 2(S0 −1)(E ∗ − µB0 ) + u∗0 (11) κ2 = = 2 1− Λ S0 (E ∗ − µB0 ) + (1/2)u∗0 € with (12) u∗0 = cIΦ∗ʹ′/B0 , S0 = 1+ (cI2Φ/Ω0 B0 ) , 2 and Ω0 =€ZeB0/Mc. The new variables W and Λ reduce to the familiar ones v /2 and λ = 2µB0/v2 in the limit of vanishing radial electric field ( u∗0 = 0 and S0 = 1). Equation (8) then reduces to the € usual result that v||2 = v2 (1 − λB / B0 ) . Equations (9) - (11) are consistent with the expressions from [1-5] written in terms of v||0, the parallel velocity evaluated at ϑ = 0. The trapped-passing boundary occurs at κ2 = 1 or€Λ = 1/(1+2ε), and is shifted and distorted from the usual boundary as noted in [1,4,5]. Equations (8) - (11) are particularly convenient for switching between ψ∗ and ψ variables. In the next section they will allow the transit averages that follow an ion trajectory to be performed holding ψ∗ fixed, while evaluating the flux surface averages at fixed ψ. € 3. Drift kinetic equation € Using ψ∗ , ϑ, E ∗ , and µ as our independent variables for an axisymmetric tokamak ˙ ∗∂f/∂ψ∗ + ϑ˙ ∂f/∂ϑ + E˙ ∗∂f/∂E ∗ = C{f} since µ is an results in the drift kinetic equation ∂f/∂t + ψ adiabatic invariant. Here f = f( ψ∗ ,ϑ, E ∗ ,µ) is the gyroaveraged ion distribution function and € ˙ ∗ = 0, allowing slow C is € the Fokker-Planck collision operator for ion-ion collisions. Using ψ ˙ = (Ze/M)∂Φ/∂t, and working to high enough order to retain E × B time dependence so that E€ ∗ € € ϑ˙ = (v||n + v d ) ⋅ ∇ϑ with n = B/B. As a result, our drift kinetic plus magnetic drifts, vd, gives € equation is simply € € + (v n + v ) ⋅ ∇ϑ∂f/∂ϑ + (Ze /M)(∂Φ /∂t)∂f/∂E = C{f} . (13) ∂f/∂t || d ∗ € dependence, (13) allows us to consider the residual zonal flow By retaining slow time problem and neoclassical transport in the pedestal in an additive fashion. To do so we let € (14) f = f∗ (ψ∗ ,E) + h(ψ∗ ,ϑ,E ∗ ,µ,t) where (15) f∗ = f∗ (ψ∗ ,E) = η(ψ∗ )[M /2πTi (ψ∗ )]3 / 2 exp[−ME /Ti (ψ∗ )] , € € with both the pseudo density, η(ψ) = ni(ψ)exp[ZeΦ(ψ)/Ti(ψ)], and the ion temperature, Ti(ψ), weakly varying functions of space in the pedestal, as required to make the entropy production vanish [11]. Then f∗ can be Taylor expanded about ψ. Notice that the ion density ni and the electrostatic potential are allowed to be strong functions of ψ so the density cannot be expanded about ψ. The background potential Φ(ψ) is assumed quadratic in ψ, while the € fluctuating potential€ is assumed to be a small correction that does not impact ion orbits. Expanding (15) about the Maxwellian € (16) fM = fM (ψ,E) = η(ψ)[M /2πTi (ψ)]3/ 2 exp[−ME /Ti (ψ)] gives ⎪⎧ Iv ⎡ ∂np i Ze ∂Φ ⎛ Mv 2 5 ⎞ ∂nTi ⎤ ⎪⎫ (17) f∗ = fM ⎨1− || ⎢ + + ⎜ − ⎟ ⎥ + ...⎬ , € ⎪⎩ Ωi ⎣ ∂ψ ⎪⎭ Ti ∂ψ ⎝ 2Ti 2 ⎠ ∂ψ ⎦ where dlnpi/dlnTi ~ LT/ρpi ~ B/Bp, with LT the ion temperature scale length. Notice that the correction to fM is f* - fM + h. Then the kinetic equation (13) to the requisite order becomes € ∂h/∂t + (v|| + u)n ⋅ ∇ϑ∂h /∂ϑ + (Ze /M)(∂Φ /∂t)∂fM/∂E = C {f − fM} , (18) € where we have allowed for the possibility of finite E × B effects by retaining the u term from ϑ˙ , and use C to denote the linearized ion-ion collision operator. For the residual zonal flow € calculation collisions are neglected. For the neoclassical transport calculations considered € linearity of (18) means that we can add the next, the time derivatives are dropped. The € neoclassical and zonal flow contributions to the perturbed ion distribution function that are found in the next two sections. 4. Neoclassical transport Dropping the time derivatives and noting that for the linearized ion-ion collision operator C {(1,v,v2/2)fM} = 0, it is convenient to define Iv ⎛ Mv 2 5 B2k ⎞ ∂nTi , (19) H = h − fM || ⎜ − + ⎟ Ωi ⎝ 2Ti 2 2〈B2 〉 ⎠ ∂ψ €where k is a flux function to be determined. The k term is added for later use to restore the C {v||fM} = 0 property when it becomes necessary to employ an approximate collision € explicit results. The B dependence of the k coefficient is made explicit to operator to obtain € obtain a form for the flows that will be divergence-free, where 〈B2 〉 = B20 with 〈...〉 denoting a flux surface average. Using (19) in the steady state version of (18) simplifies it to (20) (v|| + u)n ⋅ ∇ϑ∂h /∂ϑ = C {H} ; the form for the kinetic equation that is useful in both€the banana and€plateau regimes. 4.1 Plateau regime The plateau € regime calculation is performed in detail in [3] and easily recovered from the preceding formalism. The procedure is to first realize that because of the singular nature of H at v|| = 0, (20) is in the correct form to make the usual plateau replacement of C {H} by -νH to resolve the singularity. We then also use (5) to write the streaming operator in terms € of the ψ∗ , ϑ, E ∗ , and µ. Once the equation is solved for h the solution must be changed back to the ψ, ϑ, E, and µ variables to perform the velocity space integrals holding ψ fixed (these details of the procedure are addressed in the next subsection). The flux function k is € determined € by requiring the ions give no particle transport as required by the momentum conservation property C {v||fM} = 0 of the full linearized ion-ion collision operator. The corrected result [3] is 1+ 4U 2 + 6U 4 + 12U 6 , (21) k = J p (U 2 ) = 1+ 2U 2 + 2U 4 € with n ⋅ ∇ϑ = qR 0 , q the safety factor, R0 the major radius at ϑ = 0, and cI ∂Φ U= . (22) v iB0 ∂ψ € € € When these steps are carried out, orbit squeezing does not enter and the following results are 3 obtained for the parallel ion flow V||i = n−1 i ∫ d vv||f , radial ion heat flux 〈qi⋅ ∇ψ〉 = 〈 ∫ d 3v(Mv 2 /2)fv d€⋅ ∇ψ〉 , and bootstrap current J bs : Ip ⎡∂np i Ze ∂Φ J p (U 2 )B2 ∂nTi ⎤ (23) n iV||i = − i ⎢ + + ⎥ , MΩi ⎣ ∂ψ Ti ∂ψ 2〈B2 〉 ∂ψ ⎦ € 3/ 2 π ε2I 2p i ⎛€ Ti ⎞ ∂nTi 〈qi⋅ ∇ψ〉 = −3 L p (U 2 ) , (24) ⎜ ⎟ 2 qR 0Ω2i ⎝ M ⎠ ∂ψ € and π ε2cIBv e 2 + 4Z ⎡ ∂p ( 2 + 13Z)n e ∂Te J p (U 2 )n e ∂Ti ⎤ (25) J bs = − + ⎢ + ⎥ € 2 qR 0 ν e 〈B2 〉 Z(2 + 2Z) ⎣ ∂ψ 2( 2 + 4Z) ∂ψ 2Z ∂ψ ⎦ where pe = neTe, ve = (2Te/m)1/2, νe = 4(2π)1/2nee4lnΛ/(3m1/2Te3/2), and 1+ 4U 2 + 8U 4 + 4U 6 (4 + U 2 ) /3 (26) L p (U 2 ) = exp(−U 2 ) . 2 + 2U 4 1+ 2U € The poloidal flow of Pfirsch-Schluter trace impurities (subscript z) is then given by cIBp Ti ⎡∂np i ZTz ∂np z J p (U 2 ) ∂nTi ⎤ Vzp = − − + (27) ⎢ ⎥ . Ze〈B2 〉 ⎣ ∂ψ Z z Ti ∂ψ 2 ∂ψ ⎦ € The factors Jp and Lp account for the modifications of the usual plateau results [8] by finite radial electric field effects and equal one at U = 0. The bootstrap current is evaluated by a two-term€Laguerre expansion in combination with an adjoint method [3]. The function Jp is a monotonically increasing function approaching an asymptote of 6U2– 3+… for large U2. Therefore the poloidal ion and impurity flows are enhanced by the pedestal electric field for normal Ti profiles. The function Lp increases to a maximum of 1.46 at U2 = 0.83 before going to zero exponentially. This increase in Lp enhances the ion heat transport for normal Ti profiles for moderate U2. 4.2 Banana regime The banana regime calculation is somewhat more involved because of the need to deal with the collision operator in detail while distinguishing carefully between transit (performed at fixed ψ*) and flux surface (performed at fixed ψ) averages. The Rosenbluth potential form of the ion-ion collision operator for collisions with a background Maxwellian is used with the flux function k being determined by the need to conserve momentum in ion-ion collisions so that ion particle transport does not occur. The convenient variables to employ first are the ψ, ϑ , W and Λ variables. Only scattering normal to the trapped-passing boundary κ2 = 1 need be retained because we assume ε << 1 and neglect order ε corrections [18]. As a result, we need only evaluate the modified pitch angle scattering operator 1 ∂ ⎡ ∂ ⎛ f ⎞⎤ (28) C{f} = ⎢ jfM∇ v Λ ⋅ Q ⋅ ∇ v Λ ⎜ ⎟⎥ , € j ∂Λ ⎣ ∂Λ ⎝ fM ⎠⎦ where j = 1/∇ vW × ∇ v Λ ⋅ ∇ vϕ is the Jacobean and Q = (ν⊥ /4)(v 2 I − vv) + (ν|| /2)vv , with ν⊥ /ν i= 3(2π)1/ 2 [erf(x) − Ψ(x)]/(2x 3 ) , ν||/ν i= 3(2π)1/ 2 Ψ(x) /(2x 3 ) , x = v/vi, erf(x) the error function, Ψ(x)€= [erf(x) − xerf'(x)]/(2x 2 ) , νi the Braginski ion-ion collision frequency, and € the gyrophase ϕ giving ∇ vϕ = v⊥2 n × v . To evaluate € ∇ vW and ∇ v Λ we use (8), ΛW from (9) € (1) to obtain and (10), and ∇ vψ∗ = In /Ωi from € (1− ΛB/B0 )∇ vW = (B/B0 )W∇ v Λ + (v|| + u)n (29) € € and € € € W∇ v Λ + Λ∇ vW = [(S0B0 /B) − 2(B0 /B)(S0 −1)]v⊥ + 2(S0 −1)v . € order ε corrections the preceding give Upon neglecting j = BW/S0B0(v|| + u) and€ € € € (30) (31) ∇ v Λ = −(Λ /W)(v|| + u)n + (1− ΛB/B0 )[S0 v⊥ + 2(S0 −1)v||n + ...] ≈ −(Λ /W)(v|| + u)n . (32) The last form of ∇ v Λ is all that is needed to evaluate (28) which becomes (v + u) ∂ ⎡ ∂ ⎛ H ⎞⎤ (33) C {H} = || ⎢BΛW −2 (v|| + u)n ⋅ Q ⋅ nfM ⎜ ⎟⎥ , B ∂Λ ⎣ ∂Λ ⎝ fM ⎠⎦ where € to reduce to the usual u = 0 result we use Λ ≈ 1 since (33) is only employed for the trapped and barely passing. Equation (33) is easily written in terms of the ψ*, θ, W and Λ € by using (5), (6), and (8) - (10) so the transit averages can be performed: variables (Sv|| + u∗ ) ∂ ⎡ ∂ ⎛ H ⎞⎤ (34) C {H} = ⎢BΛW −2 (Sv|| + u∗ )n ⋅ Q ⋅ nfM ⎜ ⎟⎥ , B ∂Λ ⎣ ∂Λ ⎝ fM ⎠⎦ where ρ pi/LT ~ Bp/B << ε1/2 is assumed to neglect corrections from ∂ ψ∗/∂Λ. Notice that combining (29) and (32) give ∇ v W ≈ (v|| + u)n and W∇ v W ⋅ ∇ vΛ ≈ −Λ(v|| + u)2 . As a result, € the neglected velocity space derivatives from the full ion-ion collision operator are expected to be small as Λ derivatives acting on H are larger than W derivatives for ε << 1. To complete the specification of the collision operator we need to evaluate the coefficient n ⋅ Q ⋅ n for the trapped and barely passing ( v|| + u = Sv|| + u∗ ≈ 0 ) in both sets of variables. Using v 2 = v||2 + 2µB and the lowest order result v|| ≈ −u ≈ −u∗/S , (6) and (9) then 2 /S ≈ S µB + S u 2 , since here we can give E ∗ ≈ µB + u∗2 /2S ≈ µB + Su 2 /2 and W ≈ S0µB0 + u∗0 0 0 0 0 1/2 2 2 /S 2 , € € gives v ≈ 2W /S − u 2 ≈ 2W /S0 − u∗0 neglect ε order corrections. Using these results 0 € € which enters in ν⊥ and ν|| and allows us to write n ⋅ Q ⋅ n in both sets of variables: € € € € € 2 /2S 2 . (35) n ⋅ Q ⋅ n ≈ Wν⊥ /2S + (ν|| − ν⊥ )u 2 /2 ≈ Wν⊥ /2S0 + (ν|| − ν⊥ )u∗0 0 Recalling (20) and using the lowest order banana regime result ∂h /∂ϑ = 0 we see that the transit average of the next order version of (20) with (34) inserted for C for the passing € allows the lowest order flux function ∂h /∂Λ to be determined from ions € BC {H} =0 . (36) Sv|| + u∗ € € As usual, h = 0 for the trapped particles since the transit average is over a full bounce. Notice that the transit averages can be performed in essentially the usual manner [6-8,18]. Once ∂h /∂Λ is determined from€(36) for the passing in the ψ*, θ, W and Λ variables it is easily rewritten in the ψ, θ, W and Λ variables and the desired moments can be formed in the conventional manner. The flux function k is determined by the need to conserve momentum in ion-ion collisions and is found to be ∞ 3k 6 ⎡ 5 ∫ o dye−y (y + 2U 2 )3/2 (yν⊥ + 2U 2 ν|| ) ⎤ 2 2 (37) ≡ J b (U ) = ⎢ + U − ∞ ⎥ . 7 7 ⎣ 2 ∫ o dye−y (y + 2U 2 )1/2 (yν⊥ + 2U 2 ν|| ) ⎦ The details of this evaluation as well as those for the parallel ion flow velocity, radial ion heat flux, and bootstrap current are found in [1,2] and lead to the following expressions Ip ⎡∂np i Ze ∂Φ 7J b (U 2 )B2 ∂nTi ⎤ (38) n iV||i = − i ⎢ + − ⎥ , MΩi ⎣ ∂ψ Ti ∂ψ 6〈B2 〉 ∂ψ ⎦ 〈q i ⋅ ∇ψ〉 = −1.35 and ε νi I2 p i ∂Ti L b (U 2 ) , MΩ2i ∂ψ S (39) € J bs = −1.46 ε cIB (Z 2 + 2.21Z + 0.75) ⎡ ∂p (2.07Z + 0.88)n e ∂Te 7J b (U 2 )n e ∂Ti ⎤ − − ⎢ ⎥,(40) 〈B2 〉 6Z ∂ψ ⎦ Z(Z + 2) ⎣ ∂ψ (Z 2 + 2.21Z + 0.75) ∂ψ where 2 L b (U 2 ) = 1.53e−U ∫ o∞ dye−y (y + 2U 2 )3/2 (y + U 2 )−3/2 (y + U 2 − 5 / 2 + 7J b / 6) € ×[yerf( y + U 2 ) + (2U 2 − y)Ψ( y + U 2 )] . (41) The poloidal flow of Pfirsch-Schluter trace impurities and banana regime ions is now given by € cIBp Ti ⎡∂np i ZTz ∂np z 7J p (U 2 ) ∂nTi ⎤ Vzp = − − − (42) ⎢ ⎥ . Ze〈B2 〉 ⎣ ∂ψ Z z Ti ∂ψ 6 ∂ψ ⎦ As in the plateau regime, the normalized factors Jb and Lb are the modifications of the usual banana regime results [8] caused by finite radial electric field effects. The bootstrap current € is evaluated by inserting the Jb modification in the expressions given by Helander and Sigmar [8]. The function Jb decreases monotonically, changing sign at U2 = 1.44. As a result, the poloidal ion flow can differ importantly from the usual banana regime result. The predicted change in the flow of the background ions also impacts the impurity flow as given by (42) and is indeed observed on Alcator C-Mod [13,19]. The function Lb starts out rather flat at small U2 before going to zero exponentially. Consequently, the radial ion heat flux is € reduced by the finite electric field that acts to decrease the number of trapped and barely passing ions. 5. Zonal flow residual Analysis of the zonal flow residual is a way of considering the response to a turbulence generated, axisymmetric small amplitude, zonal flow potential fluctuation. This perturbation is assumed to have rapid radial spatial variation, but no poloidal variation: ˜ (t)exp[iG(ψ)]. The turbulence is assumed to generate the change in the potential Φ(ψ,t) = Φ in a time long compared to an ion gyro-period, but short compared to an ion bounce time, while not perturbing the density. The standard definition for the ion zonal flow residual is the ratio of the long time asymptotic value of the perturbed potential to the initial perturbation. It can be obtained by solving (18) in the collisionless limit in the absence of any neoclassical drive (f* = fM is employed since the zonal flow and neoclassical problems are additive): ˜ /∂t)fM exp(iG) . (43) ∂h/∂t + (v|| + u)n ⋅ ∇ϑ∂h /∂ϑ = (Ze /T)(∂Φ We solve (43) by assuming to to lowest order ∂h/∂ϑ = 0 and then annihilating the streaming term to next order by employing the transit average along the actual trajectory over a full € transit. Using the definition poloidal ∫ dϑA /(v|| + u)n ⋅∇ϑ ∫ dϑA /(Sv|| + u∗ )n ⋅∇ϑ A= = , (44) ∫ dϑ(v|| + u) / n ⋅∇ϑ ∫ dϑ(Sv|| + u∗ ) / n ⋅∇ϑ and periodicity in ϑ , we obtain the solution € ˜ fM exp(iG) , (45) h = (Ze /Ti )Φ € where the transit average of the eikonal G must be performed holding ψ∗ , W and Λ fixed. € ZeΦ˜ /Ti << 1 and forming the flux surface averaged perturbed ion density Assuming ˜ fM /Ti )]〉 gives n˜ i = 〈 ∫ d 3v[hexp(−iG)€ − (ZeΦ ˜ /Ti )〈 ∫ d 3vfM [exp(iG)exp(−iG) € −1]〉 . (46) n˜ i = (ZeΦ If gyroradius, as well as drift, departure polarization effects are retained then the preceding € expression becomes ˜ /Ti )〈 ∫ d 3vfM [J 0 exp(iG)J 0 exp(−iG) −1]〉 , € (47) n˜ i = (ZeΦ where the Bessel function J 0 = J 0 (k r v⊥ /Ωi ) has k r = ∇G = RBpGʹ′. For a turbulent change at t = 0 in a time much less the transit time (so the ions have gyrated many times, but not yet ˜ (t = 0) , the initial density n˜ i of (47) is simply € drifted significantly) to Φ ˜ /Ti )〈 ∫ d 3vfM [J 0 J 0 −1]〉 . € (48) n˜ i (t = 0) = (Ze€ Φ ˜ leaves the ion density unchanged and wait for Φ˜ to settle to its If we assume this change in Φ ˜ (t → ∞) , while keeping n˜ i (t → ∞) = n˜ i (t = 0) , then we can form the time asymptotic value Φ € € € (48) to obtain ratio of (47) and 〈 ∫ d 3vfM [J 0 J 0 −1]〉 € Φ(t → ∞) = €. (49) = 0) 〈 ∫ d 3vf [J exp(iG)J exp(−iG) −1]〉 Φ(t M 0 0 € € For small kr and u (49) reduces to the form of Rosenbluth and Hinton [9,10]. € € € To extend their result to finite u we need to Taylor expand G in a way that the lowest order term depends only on ψ∗ , W, Λ, and constants, and the next corrections are small in ε so that exponentials can be expanded in (49). We start by writing (50) ψ = ψ∗ − (Iu∗ /SΩi ) + [I(Sv|| + u∗ ) /SΩi ] , so the last term is a€ ε correction for the trapped and barely passing ions - the only ones of interest for our evaluation. Taking account of the poloidal variation of B in u and S as well as € ion orbit results of section 2 to retain order ε terms, we can write Ωi and using the € = Ψ + [I(Sv + u ) /SΩ ] + [4εIu /S2Ω κ 2 ][1− κ 2 sin 2 (ϑ /2)] + ... , (51) ψ ∗ || ∗ i ∗0 0 0i 2 2 where Ψ∗ ≡ ψ∗ − (Iu∗0 /S0Ω0i )[1− 4ε /S0 κ ] . We then Taylor expand G about Ψ∗ to obtain €Q − L + ... , (52) G(ψ) = G∗ + P + ... = G∗ + € where G∗ = G(Ψ∗ ) , Gʹ′∗ = Gʹ′(Ψ∗ ) , P = Q - L, Q = [I(Sv|| + u∗ ) /SΩi ]Gʹ′∗ , and € L = [4εIu∗0 /S20Ω0iκ 2 ][1− κ 2 sin 2 (ϑ /2)]G∗ʹ′ . We neglect Gʹ′ʹ′ and€ smaller terms when performing the€ Taylor expansion because the eikonal has the usual property that € ∇∇G ~ k€ small kr then gives r /a . Expanding (49) for € → ∞) Φ(t 1€ = , (53) = 0) 1+ ℜ Φ(t with € 2 2 −1 ℜ = 2n −1 ∫ d 3vfM [i(P − P) + (P 2 − 2PP + P 2 ) / 2] i 〈k r ρ i 〉 (54) to the requisite order. Noting that L ~ ε Q, then only linear terms in L need be retained in (54) along with linear and quadratic terms in Q. The evaluation then proceeds as in Landreman and Catto [5], where the full details are presented, to find ⎡ Γ(U 2 ) Λ(U,S) ⎤ € (55) ℜ = ℜ RH ⎢ +i ⎥ , 〈k r ρ pi 〉 ⎦ ⎣ S where ℜ RH = 1.6q 2 / ε is the Rosenbluth and Hinton result, ∞ € and Γ(U 2 ) = (4/3 π )exp(−U 2 ) ∫ dy(y + 2U 2 )3/2 exp(−y) 0 € ∞ ⎡ ⎤ Λ(U,S) = 2S−1/2 U ⎢SΓ(U 2 ) + 4π−1/2 (1 − S)exp(−U 2 ) ∫ dy(y + 2U 2 )1/2 exp(−y)⎥ . ⎣ ⎦ 0 (54) (54) The exponential decay of Γ and Λ for large U is due to the shift of the trapped region to the tail of the distribution function, orbit squeezing effects only enter algebraically in Λ, and the introduced by u. imaginary term is a spatial phase shift in Φ 6. Discussion We have presented a streamlined evaluation of the ion orbits and a generalized kinetic treatment that allows us to recover all the pedestal results obtained to date for neoclassical ion flow and heat flux and the bootstrap current in the banana [1,2] and plateau [3] regimes, and for the zonal flow residual in the collisionless limit [4,5]. The techniques we employ clearly distinguish between trajectory and flux surface averages and unify recent evaluations of pedestal phenomena. These modifications include the banana regime decrease and plateau regime alterations in the ion heat diffusivity, the reduction and possible reversal of the poloidal flow in the banana regime and its augmentation in the plateau regime, the increase in the bootstrap current, and the enhancement of the residual zonal flow regulation of turbulence. Acknowledgements: Work supported by U. S. Department of Energy grants at DE-FG0291ER-54109 at MIT and DE-AC52-06NA-25396 at LANL, and the European Communities under Association Contract between EURATOM and Vetenskapsradet. PJC is grateful for the hospitality of the Isaac Newton Institute for Mathematical Sciences in Cambridge, England while preparing portions of this article. References [1] Kagan G. and Catto P.J. 2010 Plasma Phys. Control. Fusion 52 055004 and 079801 [2] Kagan G. and Catto P.J. 2010 Phys. Rev. Lett. 102 045002 [3] I. Pusztai and P. J. Catto 2010 Plasma Phys. Control. Fusion 52 075016 [the corrected result for J is as given here, an errata is being prepared] [4] Kagan G. and Catto P.J. 2009 Phys. Plasmas 16 056105 and 099902 [5] Landreman M. and Catto P.J. 2010 Plasma Phys. Control. Fusion 52 085003 [6] Hinton F.L. and Hazeltine R.D. 1976 Rev. Mod. Phys. 48 239 [7] Hirshman S.P. and Sigmar D.J. 1981 Nucl. Fusion 21 1079 [8] Helander P. and Sigmar D.J. 2002 Collisional Transport in Magnetized Plasmas (Cambridge Monographs on Plasma Physics) ed M G Haines et al (Cambridge: Cambridge University Press) [9] Rosenbluth M.N. and Hinton F.L. 1998 Phys. Rev. Lett. 80 724 [10] Xiao Y., Catto P.J. and Dorland W. 2007 Phys. Plasmas 14 055910 [11] Kagan G. and Catto P.J. 2008 Plasma Phys. Control. Fusion 50 085010 [12] McDermott R., Lipschultz B., Hughes J.W., Catto P.J., Hubbard A.E., Hutchinson I.H., Greenwald M., LaBombard B., Marr K., Reinke M.L., Rice J.E., Whyte D. and Alcator CMod Team 2009 Phys. Plasmas 16 056103 [13] Marr K.D., Lipschultz B., Catto P.J., McDermott R.M., Reinke M.L. and Simakov A.N. 2010 Plasma Phys. Control. Fusion 52 055010 [14] DeGrassie J. 2008 private communication, General Atomics [15] Kovrizhnikh L.M. 1969 Zh. Eksp. Teor. Fiz. 56 877 [1969 Sov. Phys. JETP 29 475] [16] Rutherford P.H. 1970 Phys. Fluids 13 482 [17] Hazeltine R.D. 1989 Phys. Fluids B 1 2031 [18] Rosenbluth M.N., Hazeltine R.D. and Hinton F.L. 1972 Phys. Fluids 15 116 [19] Kagan G., Marr K.D., Catto P.J., M. Landreman, Lipschultz B. and R. McDermott, to be submitted to PPCF.